[
https://issues.apache.org/jira/browse/KAFKA-5261?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16146057#comment-16146057
]
ASF GitHub Bot commented on KAFKA-5261:
---------------------------------------
GitHub user simplesteph opened a pull request:
https://github.com/apache/kafka/pull/3756
KAFKA-5261 added cached authorizer
attempt to improve Kafka performance by caching the results of the
authorisation calls.
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/simplesteph/kafka KAFKA-5261
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/kafka/pull/3756.patch
To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:
This closes #3756
----
commit 4fc2c0f365b675eb471a848b29001973ec1c2aa5
Author: Stephane Maarek <[email protected]>
Date: 2017-08-29T20:03:56Z
added cached authorizer
----
> Performance improvement of SimpleAclAuthorizer
> ----------------------------------------------
>
> Key: KAFKA-5261
> URL: https://issues.apache.org/jira/browse/KAFKA-5261
> Project: Kafka
> Issue Type: Improvement
> Affects Versions: 0.10.2.1
> Reporter: Stephane Maarek
>
> Currently, looking at the KafkaApis class, it seems that every request going
> through Kafka is also going through an authorize check:
> {code}
> private def authorize(session: Session, operation: Operation, resource:
> Resource): Boolean =
> authorizer.forall(_.authorize(session, operation, resource))
> {code}
> The SimpleAclAuthorizer logic runs through checks which all look to be done
> in linear time (except on first run) proportional to the number of acls on a
> specific resource. This operation is re-run every time a client tries to use
> a Kafka Api, especially on the very often called `handleProducerRequest` and
> `handleFetchRequest`
> I believe a cache could be built to store the result of the authorize call,
> possibly allowing more expensive authorize() calls to happen, and reducing
> greatly the CPU usage in the long run. The cache would be invalidated every
> time a change happens to aclCache
> Thoughts before I try giving it a go with a PR?
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
